Welcome!

Linux Containers Authors: Elizabeth White, AppDynamics Blog, Christopher Keene, Janakiram MSV, Greg O'Connor

Related Topics: Linux Containers

Linux Containers: Article

VM: Virtual Memory or Virtual Mayhem?

How to solve the virtual memory problem in all Linux kernels.

(LinuxWorld) -- The fur has been flying in the Linux kernel development community of late, particularly because there's a lot of contention about how Linux should do virtual memory management (VM). Andrea Arcangeli and Rik van Riel have the two most visible opinions on how the VM should work, so they're often seen as at the heart of the struggle. However, it goes much deeper than that. The VM even seems like a hot button between Linus Torvalds and Alan Cox, although I'm guessing they're not as passionate about the issue as the journalists would like them to be to get mileage out of the topic.

Take us, for example. None other than our own Joshua Drake called the 2.4 Kernel the Kernel of pain, primarily because of problems with the VM algorithms and changes. Yours truly complained a little about the VM last August.

I'm about to weigh in once again about how I feel about VM, only in more detail this time. We're definitely into milking this topic for all it's worth.

After looking at the various VM algorithms in play, I've come up with an alternative. I'm sure Andrea Arcangeli and Rik van Riel are flooded with the same old suggestions over and over again, and it's entirely possible they've heard this one before. Nevertheless, I'll propose it, anyway.

First, let me walk you through a little of the history that gave me the notion of how best to deal with the Linux VM issue.

The brain-dead system that wouldn't die

Back when I was a programmer by trade, one of the projects I worked on involved Fortran programming for a farm of PDP-11s. I honestly don't remember which OS we had installed on the PDPs, but if I had to guess I'd say it was probably RSX-11/M.

What I'll never forget, however, is the one quirk I hated beyond all others. You couldn't run executable files unless they were stored in contiguous blocks on the disk.

Programmers are a virtual disk frag factory, so we ran out of contiguous space almost daily. This means we often found ourselves in the uncomfortable situation where we could no longer test the modifications to our programs until someone defragged the drives. The worst of it was that, if there was such a thing as a disk defrag utility available, the systems guys obviously didn't know about it. They solved the problem by taking the systems off line, backing up, and then restoring the contents of the drives. We couldn't even edit code while we waited.

By the way, Microsoft credits Windows NT architect Dave Cutler as the designer of RSX-11/M (see resources). As tempting as it might be to compare the brain-dead pieces of Windows NT to the bizarre behavior of RSX-11/M, it would be unfair to do so for two reasons. First, more reliable sources than Microsoft only credit Cutler for the design of VMS. Cutler borrowed some RSX-11/M code for VMS, but didn't design RSX-11/M. Second, RSX-11/M was probably only brain-dead because the PDP-11 wasn't much of a brain.

Any computer historian knows the PDP-11 was a breakthrough in affordable computing in its time. Compared to today's desktops, it would be like taking out a second mortgage to buy an abacus. A PDP-11 computer with 8K words of core memory and a 256K disk cost about $30,000 in 1972. It could run you thousands of dollars for an additional 4K of memory.

It may be possible that the RSX-11/M architect was an idiot. It is also possible that the resident portion of the RSX-11/M kernel would have to exceed the size of the typical memory configuration for the PDP-11 in order to change this behavior, which would inflate the price of the system in the process.

I don't happen to recall how much memory we had on our PDP-11s or how much of it was used by RSX-11/M. However, I do know we never had enough. I had to implement overlays to make my Fortran programs work. In case you're not an old fart as I am, overlays are a bit like implementing virtual memory at the application level instead of the OS. You split off parts of your program into modules called overlays. The main program loads an overlay whenever it needs the functions in that overlay.

The trick is to make sure that none of your overlays rely on functions that reside in any other overlays, because your goal is to have only one overlay in memory at a time. I'm not even sure it was possible to have more than one loaded, but it's been too long to remember. Regardless, I'd bet the company paid at least double my yearly salary for some of those PDP-11s, so it was worth the relatively minimal effort for me to break up my program into overlays.

The new VM proposal

What does this have to do with virtual memory performance on the Linux 2.4 kernel?

Everything.

First, think about what the VM is for. The VM is like an extremely sophisticated automated overlay system, only it deals with many more types of memory storage. It is the part of the OS that comes to the rescue when you have used up all your expensive memory and need more.

Here's how it works. The OS finds some data in memory that can safely be removed and stores it to the cheaper storage (disk swap space), which frees up some expensive memory for other use. When a program needs the data that has been swapped to disk, the OS swaps something else out to disk and brings the needed data back into memory.

Most of the arguments about the VM in Linux revolve around how the OS should decide which memory is swapped and when, and methods to make the process fast and painless.

The remaining arguments are usually about what the OS needs to do when you've filled up all available memory and swap space but some task still needs even more memory. In this case, most people agree that the OS needs to kill one or more running tasks to free up memory. Since you're talking about stopping programs dead in their tracks, you have to address the issue of how the OS decides which tasks are less important than others and can afford to be killed.

If you don't see how this relates to my PDP story yet, then here's a hint. The controversy over the VM involves several extremely talented programmers, each of whom could command hundreds of thousands of dollars per year in salaries. They have been spending a great deal of their time and brain power over the past years figuring out how to squeeze the best performance out of systems with limited RAM and drive space.

Still don't get it? Then let me get right to my proposal for a new VM algorithm for Linux. Granted, this VM algorithm is not meant for typical system loads, but it would solve the most annoying VM problems.

I propose that we create a kernel daemon that checks for either of the following two conditions:

  1. The system swaps to disk so much that you see a severe degradation in performance.
  2. A task needs memory after all available RAM and swap is filled.

If either condition is met, the kernel then kills all tasks except those it needs to display the following message on the screen: "Lay off the doughnuts this week and spend the money to buy another DIMM, you penny pinching skinflint!"

As a bonus, the daemon could check the Internet for current pricing and replace the part about "doughnuts" with some comparison that better represents the current state of the market.

Again, I apologize to Andrea and Rik if this has already been suggested, and I suspect it has. Nevertheless, it was therapeutic, if not useful, to offer the advice.

More Stories By Nicholas Petreley

Nicholas Petreley is a computer consultant and author in Asheville, NC.

Comments (0)

Share your thoughts on this story.

Add your comment
You must be signed in to add a comment. Sign-in | Register

In accordance with our Comment Policy, we encourage comments that are on topic, relevant and to-the-point. We will remove comments that include profanity, personal attacks, racial slurs, threats of violence, or other inappropriate material that violates our Terms and Conditions, and will block users who make repeated violations. We ask all readers to expect diversity of opinion and to treat one another with dignity and respect.


@ThingsExpo Stories
With so much going on in this space you could be forgiven for thinking you were always working with yesterday’s technologies. So much change, so quickly. What do you do if you have to build a solution from the ground up that is expected to live in the field for at least 5-10 years? This is the challenge we faced when we looked to refresh our existing 10-year-old custom hardware stack to measure the fullness of trash cans and compactors.
The emerging Internet of Everything creates tremendous new opportunities for customer engagement and business model innovation. However, enterprises must overcome a number of critical challenges to bring these new solutions to market. In his session at @ThingsExpo, Michael Martin, CTO/CIO at nfrastructure, outlined these key challenges and recommended approaches for overcoming them to achieve speed and agility in the design, development and implementation of Internet of Everything solutions wi...
Cloud computing is being adopted in one form or another by 94% of enterprises today. Tens of billions of new devices are being connected to The Internet of Things. And Big Data is driving this bus. An exponential increase is expected in the amount of information being processed, managed, analyzed, and acted upon by enterprise IT. This amazing is not part of some distant future - it is happening today. One report shows a 650% increase in enterprise data by 2020. Other estimates are even higher....
Today we can collect lots and lots of performance data. We build beautiful dashboards and even have fancy query languages to access and transform the data. Still performance data is a secret language only a couple of people understand. The more business becomes digital the more stakeholders are interested in this data including how it relates to business. Some of these people have never used a monitoring tool before. They have a question on their mind like “How is my application doing” but no id...
The 19th International Cloud Expo has announced that its Call for Papers is open. Cloud Expo, to be held November 1-3, 2016, at the Santa Clara Convention Center in Santa Clara, CA, brings together Cloud Computing, Big Data, Internet of Things, DevOps, Digital Transformation, Microservices and WebRTC to one location. With cloud computing driving a higher percentage of enterprise IT budgets every year, it becomes increasingly important to plant your flag in this fast-expanding business opportuni...
Identity is in everything and customers are looking to their providers to ensure the security of their identities, transactions and data. With the increased reliance on cloud-based services, service providers must build security and trust into their offerings, adding value to customers and improving the user experience. Making identity, security and privacy easy for customers provides a unique advantage over the competition.
19th Cloud Expo, taking place November 1-3, 2016, at the Santa Clara Convention Center in Santa Clara, CA, will feature technical sessions from a rock star conference faculty and the leading industry players in the world. Cloud computing is now being embraced by a majority of enterprises of all sizes. Yesterday's debate about public vs. private has transformed into the reality of hybrid cloud: a recent survey shows that 74% of enterprises have a hybrid cloud strategy. Meanwhile, 94% of enterpri...
Smart Cities are here to stay, but for their promise to be delivered, the data they produce must not be put in new siloes. In his session at @ThingsExpo, Mathias Herberts, Co-founder and CTO of Cityzen Data, will deep dive into best practices that will ensure a successful smart city journey.
SYS-CON Events announced today that 910Telecom will exhibit at the 19th International Cloud Expo, which will take place on November 1–3, 2016, at the Santa Clara Convention Center in Santa Clara, CA. Housed in the classic Denver Gas & Electric Building, 910 15th St., 910Telecom is a carrier-neutral telecom hotel located in the heart of Denver. Adjacent to CenturyLink, AT&T, and Denver Main, 910Telecom offers connectivity to all major carriers, Internet service providers, Internet backbones and ...
There is growing need for data-driven applications and the need for digital platforms to build these apps. In his session at 19th Cloud Expo, Muddu Sudhakar, VP and GM of Security & IoT at Splunk, will cover different PaaS solutions and Big Data platforms that are available to build applications. In addition, AI and machine learning are creating new requirements that developers need in the building of next-gen apps. The next-generation digital platforms have some of the past platform needs a...
SYS-CON Events announced today Telecom Reseller has been named “Media Sponsor” of SYS-CON's 19th International Cloud Expo, which will take place on November 1–3, 2016, at the Santa Clara Convention Center in Santa Clara, CA. Telecom Reseller reports on Unified Communications, UCaaS, BPaaS for enterprise and SMBs. They report extensively on both customer premises based solutions such as IP-PBX as well as cloud based and hosted platforms.
I wanted to gather all of my Internet of Things (IOT) blogs into a single blog (that I could later use with my University of San Francisco (USF) Big Data “MBA” course). However as I started to pull these blogs together, I realized that my IOT discussion lacked a vision; it lacked an end point towards which an organization could drive their IOT envisioning, proof of value, app dev, data engineering and data science efforts. And I think that the IOT end point is really quite simple…
Internet of @ThingsExpo, taking place November 1-3, 2016, at the Santa Clara Convention Center in Santa Clara, CA, is co-located with 19th Cloud Expo and will feature technical sessions from a rock star conference faculty and the leading industry players in the world. The Internet of Things (IoT) is the most profound change in personal and enterprise IT since the creation of the Worldwide Web more than 20 years ago. All major researchers estimate there will be tens of billions devices - comp...
DevOps at Cloud Expo, taking place Nov 1-3, 2016, at the Santa Clara Convention Center in Santa Clara, CA, is co-located with 19th Cloud Expo and will feature technical sessions from a rock star conference faculty and the leading industry players in the world. The widespread success of cloud computing is driving the DevOps revolution in enterprise IT. Now as never before, development teams must communicate and collaborate in a dynamic, 24/7/365 environment. There is no time to wait for long dev...
Data is the fuel that drives the machine learning algorithmic engines and ultimately provides the business value. In his session at Cloud Expo, Ed Featherston, a director and senior enterprise architect at Collaborative Consulting, will discuss the key considerations around quality, volume, timeliness, and pedigree that must be dealt with in order to properly fuel that engine.
Pulzze Systems was happy to participate in such a premier event and thankful to be receiving the winning investment and global network support from G-Startup Worldwide. It is an exciting time for Pulzze to showcase the effectiveness of innovative technologies and enable them to make the world smarter and better. The reputable contest is held to identify promising startups around the globe that are assured to change the world through their innovative products and disruptive technologies. There w...
Personalization has long been the holy grail of marketing. Simply stated, communicate the most relevant offer to the right person and you will increase sales. To achieve this, you must understand the individual. Consequently, digital marketers developed many ways to gather and leverage customer information to deliver targeted experiences. In his session at @ThingsExpo, Lou Casal, Founder and Principal Consultant at Practicala, discussed how the Internet of Things (IoT) has accelerated our abil...
Is the ongoing quest for agility in the data center forcing you to evaluate how to be a part of infrastructure automation efforts? As organizations evolve toward bimodal IT operations, they are embracing new service delivery models and leveraging virtualization to increase infrastructure agility. Therefore, the network must evolve in parallel to become equally agile. Read this essential piece of Gartner research for recommendations on achieving greater agility.
SYS-CON Events announced today that Venafi, the Immune System for the Internet™ and the leading provider of Next Generation Trust Protection, will exhibit at @DevOpsSummit at 19th International Cloud Expo, which will take place on November 1–3, 2016, at the Santa Clara Convention Center in Santa Clara, CA. Venafi is the Immune System for the Internet™ that protects the foundation of all cybersecurity – cryptographic keys and digital certificates – so they can’t be misused by bad guys in attacks...
For basic one-to-one voice or video calling solutions, WebRTC has proven to be a very powerful technology. Although WebRTC’s core functionality is to provide secure, real-time p2p media streaming, leveraging native platform features and server-side components brings up new communication capabilities for web and native mobile applications, allowing for advanced multi-user use cases such as video broadcasting, conferencing, and media recording.